Udacity – Deep Learning Nanodegree Foundation – Week 1

I am in the first week of the course of the Deep Learning Nanodegree and the main topics are around developing an intuition around feedforward neural networks, backpropagation (feeding errors back into the network to improve it), the sigmoid function, and how AND, OR, and NOT logic is performed in neural networks. Additionally, Anaconda and Jupyter Notebooks are all lightly covered.

I am finding that the biggest challenge is implementing backpropagation by hand in the first project – “bike sharing” – that makes use of the use UCI bike sharing dataset.

I have to manually draw out the “flow” of outputs, inputs, application of weights, updating of weights, calculating of gradients, etc. Still working on getting all of the unit tests to pass for the first project…